Meta’s Muse has done in twelve days what most consumer apps never do at all: it beat ChatGPT at its own game, on its own turf, in the U.S. App Store. The personal AI agent, which Meta calls “the world’s first personal AI agent,” launched on September 8, 2026, and by September 21 had climbed past 2.5 million downloads, according to Sensor Tower data reported by CNBC. Reuters put the two-week figure closer to 2.8 million. Either way, Muse became the top free app on Apple’s App Store on September 18, dislodging ChatGPT from a perch it had held for years.
The download curve is the headline, but the story underneath it is about what Meta actually built, how fast rivals can respond, and whether a free-tier land grab can turn into a durable product. This piece breaks down the adoption numbers, the model behind the app, the market reaction, and where the AI agent race goes from here.
Muse’s download numbers, day by day
Muse shipped in the United States on iOS, Android, and through muse.ai on September 8, then expanded to Canada on September 18. In its first five days it logged roughly 730,000 downloads, Sensor Tower told CNBC. Growth didn’t slow from there. Reuters cited an average daily download growth rate of 55% across the first ten days, and Meta’s app reportedly hit a single-day record of 264,000 U.S. downloads on September 19, its third straight day above 200,000.
By September 18, Muse had an estimated 448,000 daily active users, just ten days after launch. Apptopia data cited by the Motley Fool put the 12-day download tally at 1.8 million iOS downloads across the U.S. and Canada, while a separate estimate cited by Reuters and Investing.com placed the combined iOS-and-Android total near 2.8 million over roughly the same window. The split, per Sensor Tower’s September 21 snapshot: about 1.5 million iOS installs and 1.1 million on Android.
| Milestone | Downloads / Users | Source |
|---|---|---|
| First 5 days (Sept. 8-13) | ~730,000 downloads | Sensor Tower, via CNBC |
| Day 10 daily active users | ~448,000 DAU | Sensor Tower |
| Single-day peak (Sept. 19) | 264,000 U.S. downloads | Sensor Tower, via Investing.com |
| 12 days (iOS only, US/Canada) | ~1.8 million | Apptopia, via Motley Fool |
| By Sept. 21 (all platforms) | ~2.5 million | Sensor Tower, via CNBC |
| Two weeks (all platforms) | ~2.8 million | Sensor Tower, via Reuters |
Note that different trackers use different windows and count Android differently, so the totals don’t line up perfectly. What every source agrees on is direction: Muse’s curve is steep, and it topped the free-app charts on both Apple and Google’s storefronts inside two weeks.
How the launch stacks up against ChatGPT, Claude, and Grok
Apptopia’s comparison, cited by the Motley Fool, is the clearest cross-product benchmark so far: Muse’s first 12 days outpaced ChatGPT’s first 12 days after its November 2022 debut, when OpenAI’s chatbot logged about 1.3 million downloads in the same stretch. That’s a rough comparison across very different eras of app-store behavior and smartphone penetration, but it’s the one analysts keep repeating because it’s the only precedent big enough to matter.
Muse also cleared its direct 2026 rivals by a wide margin during a comparable early window: Anthropic’s Claude reportedly drew around 400,000 downloads and xAI’s Grok around 200,000 in equivalent periods, per the figures Reuters and CNBC gathered from Sensor Tower and Apptopia. Neither company disputed those figures directly.
| Product | Maker | Reported downloads (comparable early window) |
|---|---|---|
| Muse | Meta | ~1.8-2.8 million (12-14 days) |
| ChatGPT (Nov. 2022 launch) | OpenAI | ~1.3 million (first 12 days) |
| Claude app | Anthropic | ~400,000 (comparable window) |
| Grok app | xAI | ~200,000 (comparable window) |
The gap matters because Meta didn’t win on model quality alone. It won on distribution. Muse ships as a standalone app but is designed to plug into WhatsApp and Instagram, two networks with billions of existing users who never had to search an app store to find it.
What Muse actually does
Muse is built to take a goal, not a prompt, and run with it. Meta says the agent can send emails, book travel, shop online, schedule appointments, and buy movie tickets on a user’s behalf, either through the dedicated app or inside Meta’s messaging products. That framing, an agent that acts rather than just answers, is why Meta calls it a personal agent instead of a chatbot.
Meta has also said Muse will eventually run on Ray-Ban Meta glasses, extending the same task-completion model to a voice-first, hands-free device. That’s a distribution channel none of Muse’s rivals currently own outright, and it’s a big part of why analysts treat the smart-glasses roadmap as more than a hardware footnote.
Muse Spark: the model behind the agent
Muse runs on Muse Spark, which Meta describes as its most capable model to date. Reporting has specifically named Muse Spark 1.3 as the version powering the agent at launch, a model Meta has separately positioned as a step up in tool-calling reliability, a detail covered in earlier reporting on Muse Spark 1.3’s coding and tool-use gains. Meta hasn’t published a full architecture paper for Spark, so exact parameter counts and training details remain undisclosed.
What is confirmed is the pricing structure wrapped around the model. Meta offers a free tier that includes 100 million tokens per week, alongside paid plans at $20 and $100 a month depending on usage. Chief AI officer Alexandr Wang described that free-to-paid ladder as the mechanism meant to convert casual downloads into recurring usage.
Why Zuckerberg is calling this a bet on “personal superintelligence”
Mark Zuckerberg has framed Muse as more than a product launch. “In the coming years, I expect that Muse is going to grow into the personal superintelligence that billions of people around the world are going to use to accomplish their goals and improve their lives,” Zuckerberg said, according to CNBC.
He’s described the ambition in broader terms elsewhere too: “Everyone will have an exceptionally capable personal agent that understands you, your goals, and everything you care about. Your agent will work 24/7 on your behalf to improve your relationships, health, career, finances, home management, hobbies, and more,” Zuckerberg said, as reported by Tech Insider.
Meta CTO Andrew Bosworth has been more specific about where he thinks the company’s edge actually sits, arguing it’s not just model quality. “There are a bunch of pieces you have to connect it to: product, distribution, and the consumer experience. It’s the collection of all four that I think is our advantage relative to competitors, most of whom, whether it’s Apple, Anthropic, OpenAI, Google, only have one of those pieces,” Bosworth told Big Technology.
The stock market’s reaction
Wall Street moved fast. Meta shares jumped more than 11% on September 21 after Wells Fargo raised its price target and Muse hit the top of Apple’s App Store, according to Yahoo Finance. Other outlets tracked the broader run differently: CNBC described the stock as up more than 20% since the launch, and a separate calculation tied to trading data through September 24 put the gain closer to 26.8%, rising from $613.48 to $777.59. Investors.com reported a 30% climb over the trailing month as Muse kept climbing the download charts.
Those figures don’t all measure the same window, so treat them as directional rather than a single verified number. What’s consistent across every source is that Meta’s stock move tracked the download curve closely enough that traders started treating app-store rank as a leading indicator for the earnings call.
Why a venture capitalist says the free tier is the real story
Analysts keep circling back to distribution rather than raw model benchmarks. One venture capitalist put it plainly to Yahoo Finance: “Meta can put Muse in front of billions of people through WhatsApp and Instagram, and with a free base tier, broad adoption is quite likely,” according to Yahoo Finance.
That view lines up with what Bret Taylor, chairman of Sierra and former co-CEO of Salesforce, has said about the shift toward agents generally, independent of Meta specifically. “In 2026, every company needs an AI agent. And the vast majority of the digital interactions you’ll have with your customers will be via your agent,” Taylor said on a podcast interview. Muse’s early numbers read like an early data point for that thesis, even if Taylor wasn’t speaking about Meta’s app directly.
Historical context: this isn’t Meta’s first swing at personal AI
Meta has tried versions of a personal assistant before, from Meta AI’s chatbot rollout across its apps to the smart-glasses voice assistant baked into Ray-Ban Meta hardware. None of those efforts produced a chart-topping standalone app. What changed with Muse is scope: it’s built around completing multi-step tasks rather than answering questions, and it launched with a free tier generous enough (100 million tokens a week) to remove the biggest early barrier to trying it.
The comparison to ChatGPT’s November 2022 debut is the one that keeps surfacing in coverage, and for good reason. ChatGPT needed roughly two months to cross a million users, largely through word of mouth with no dedicated mobile app for its first several months. Muse cleared a larger download number in less than two weeks, with a purpose-built app on day one and access through Meta’s messaging platforms.
Not everyone is on board
Muse’s rise hasn’t been friction-free. Amazon has restricted the agent’s access over concerns tied to how it interacts with shopping and account data, an issue detailed in earlier reporting on Amazon’s pushback against Muse. That kind of platform-level resistance is a preview of the fights every agent maker will likely face as agents start acting on users’ behalf across services that weren’t built to be automated.
There’s also a sustainability question sitting underneath the download numbers. App-store chart position rewards novelty. Meta AI’s earlier chatbot integrations saw strong initial curiosity that didn’t always convert into habitual daily use. Whether Muse’s 448,000 daily active users at day 10 keeps compounding, or plateaus the way many viral consumer apps do, is the number that will actually determine whether this becomes a durable product line or a strong launch week.
How rivals are likely to respond
OpenAI, Anthropic, and xAI aren’t standing still. OpenAI has continued pushing pricing and capability on its own model lineup, including the GPT-6 Sol and Luna launch, which undercut Claude on price. xAI has leaned on aggressive pricing too, with Grok 4.7 arriving at rates designed to squeeze coding-focused rivals. None of that pricing pressure targets Muse’s core advantage directly, though, because Muse’s edge isn’t its API price sheet. It’s that Meta can put the product in front of WhatsApp and Instagram users without anyone opening an app store search bar.
That distribution gap is why Bosworth’s four-piece framework, model, product, distribution, and consumer experience, keeps coming up in coverage. Apple has models but limited agent distribution outside its own hardware. Anthropic and OpenAI have strong models and growing enterprise footprints but nothing like Meta’s two-billion-plus messaging user base. Google has both scale and models but hasn’t shipped a standalone consumer agent app with Muse’s early velocity.
Market impact beyond Meta’s stock price
Muse’s launch fed into a broader rally across AI-adjacent stocks in late September, with chipmakers and data-center suppliers catching a bid alongside Meta, according to Investing.com’s coverage of the “tech rally” tied to Muse’s App Store run. That’s a familiar pattern: any signal that consumer AI demand is real and growing tends to lift the entire supply chain behind it, from GPU vendors to cloud infrastructure providers, not just the company that shipped the product.
It also puts pressure on Meta’s own AI narrative. The company has spent heavily building out AI infrastructure and talent, including Alexandr Wang’s move into a top AI role, and Muse is the first consumer product to test whether that spending translates into a product people actually open every day rather than a research showcase.
What to watch next
Three numbers will decide whether Muse’s launch surge becomes a real product story: retention past the first month, the ratio of free-tier users converting to the $20 or $100 paid plans, and whether Meta expands availability beyond the U.S. and Canada. None of those figures are public yet. Meta’s next earnings call is the most likely venue for the company to disclose retention or conversion data, if it chooses to at all.
Predictions: where this goes from here
- Meta will expand Muse beyond the U.S. and Canada within the next two to three months, most likely starting with English-speaking markets where WhatsApp penetration is already high.
- Daily active user growth will slow from its current pace as the novelty effect fades, mirroring the pattern seen with most viral consumer AI launches since ChatGPT.
- Expect at least one more platform-level dispute similar to Amazon’s restrictions, as retailers and service providers push back on agents acting on behalf of users inside their apps.
- Rivals won’t try to out-download Muse directly. Instead, look for OpenAI, Anthropic, and Google to lean harder into enterprise and developer channels where Meta has less reach.
- Ray-Ban Meta glasses integration will become the next major Muse announcement, likely timed to a hardware event, since it’s the one distribution edge no competitor can copy quickly.
The bottom line
Muse’s first two weeks prove Meta can win a download race when it wants to. Free tokens, a built-in audience across WhatsApp and Instagram, and a task-completion pitch got it there faster than ChatGPT managed in 2022. The harder test starts now: turning a chart-topping app into daily habit, converting free users into paying ones, and handling the platform friction, like Amazon’s restrictions, that comes with an agent that acts on a user’s behalf across services Meta doesn’t control.
Related
- Meta Muse Spark 1.3 Cuts Tool Calls 20%, Tokens 25% [2026]
- Amazon Blocks Meta’s Muse Agent Over 3 Concerns [2026]
- Grok 4.7 Ships at $2/$6, Trails GPT-6 by 34 Points [2026]
- Claude Opus 5.5 Beats GPT-5.6 Sol at Third the Cost [2026]
- GPT-6 Sol, Luna Launch at 50% Off, Undercut Claude [2026]
FAQ
How many downloads does Meta’s Muse have?
Estimates range from about 2.5 million to 2.8 million downloads within two weeks of its September 8, 2026 launch, according to Sensor Tower data reported by CNBC and Reuters. Apptopia’s separate tracking put the 12-day iOS total at roughly 1.8 million across the U.S. and Canada.
What model powers Meta’s Muse?
Muse runs on Muse Spark, which Meta calls its most capable model to date. Reporting has specifically named Muse Spark 1.3 as the version powering the agent at launch.
How much does Muse cost?
Muse offers a free tier with 100 million tokens per week, plus paid plans at $20 or $100 a month depending on usage, according to Meta’s announcement.
Where is Muse available?
Muse launched in the United States on iOS, Android, and muse.ai on September 8, 2026, and expanded to Canada on September 18, 2026.
Did Muse really beat ChatGPT to the top of the App Store?
Yes. CNBC reported that Muse overtook OpenAI’s ChatGPT as the top free iOS app in the U.S. on September 18, 2026, and it was also reported at the top of Google Play’s U.S. rankings around the same time.
How much did Meta’s stock rise after Muse launched?
Reported figures vary by source and time window: Yahoo Finance reported an 11% single-day jump on September 21, while other outlets cited gains ranging from roughly 20% to 30% measured over longer periods since the launch. There’s no single officially confirmed percentage tied only to Muse.
Will Muse work with Ray-Ban Meta glasses?
Meta has said Muse will eventually be available through Ray-Ban Meta glasses, though the company hasn’t given a specific date for that rollout.
How does Muse compare to Anthropic’s Claude and xAI’s Grok in downloads?
Reuters and CNBC cited figures showing Muse’s early downloads far exceeded both apps in comparable early windows, with Claude at roughly 400,000 and Grok at roughly 200,000 downloads against Muse’s 1.8 million-plus over a similar stretch.
